Distributed network data storage system and method

Abstract

The present invention is a system and method for distributed, highly scalable, wide area peer-to-peer network data storage. The functionally equivalent servers in the system are divided into groups. Each server maintains a dynamic list which is polled to determine the availability of the closest neighbor servers. Each server is switched between the groups of servers to optimize network connectivity parameters. Data and directory files are divided into a plurality of pieces which are stored on different servers. Files are uniformly and independently named, utilizing a tree with a common root, logical pathways, and unique file identifiers. When a server receives a client request for file system access, the plurality of file pieces are collected and sent to the client server from the neighbor servers simultaneously in order to optimize bandwidth. Servers with maximum throughput capacity are utilized for highest transmission speed and reduced processing time.

Description

REFERENCE TO RELATED APPLICATIONS[0001] This application claims the benefit of U.S. Provisional Application for Patent No. 60 / 282,468 titled "Distributed Highly Scalable Wide Area Peer To Peer Network Data Storage With Uniform Name Space And Self Optimized Delivery Model" filed on Apr. 9, 2001 for priority under 35 U.S.C. .sctn.119(e), is related thereto, is commonly assigned therewith, and the subject matter thereof is incorporated herein by reference in its entirety.[0002] 1. Field of the Invention[0003] The present invention pertains to distributed data storage; more particularly, the present invention describes a distributed, highly scalable, wide area peer-to-peer network data storage system and method.[0004] 2. Background of the Invention[0005] The problem of data storage in a computer network appeared immediately after computers started to be connected to one another. The data storage problem has typically been solved by providing one of the computers with the necessary servi...

AI Technical Summary

Problems solved by technology

The problem of data storage in a computer network appeared immediately after computers started to be connected to one another.

Writing in the same file can lead to content errors.

The next problem is that the data storage file itself, being situated on only one server (and more importantly, on one hard disk), cannot be accessed in case of failures in the data storage file server equipment, network problems, or server software failures (i.e., many "failure points" appear, which means "no access" to the data).

Such inability to access data even appears in case of workable server equipment when a computer user must reboot an operating system.

When rebooting a data storage file, access service becomes unavailable for the clients, which means a break in use again.

Particularly, it meant the absence of the clusterization process support software and the associated mistakes.

Nevertheless, the equipment failure of clusterization hardware led to the problem of service unavailability.

More problems appear if the server and its data storage services are separated by distance and are connected via the internet.

The number of client computers can be so large that the server itself can become inaccessible when trying to respond to the client computers' requests for data (due to the insufficient network throughput capacity or server's inability to answer every request).

That is why it is clear that the approach with one stand-alone server (and even a group of servers located in one place) and client computers connected via the Internet leads to failures caused by the distance between the server and the client.

Such a decision at the same time presents many additional problems, such as data synchronization.

Presently, these solutions typically use very complicated algorithms to access stored data.

Network distributed file systems are arranged in a more complicated way.

Its loss entails disappearance of access to certain files.

Thus, there exists a potential interchangeability of files, but, even if properly organized, faulttolerance level of the system is not higher than usual.

If there is no connection with the data storage file server, the AFS system doesn't usually make it possible to use the data storage file.

Nevertheless, such an approach can lead to problems if several clients working with the same file simultaneously make concurrent changes that can lead to incorrect work in the future.

The presence of many copies of different file modifications complicates file system serviceability control, i.e., data coherence.

Access to files not in cache means that the support of the server system where different data is stored on different servers can lead to access problems in the case where connection to the server is lost.

The main problem with such a method is the necessity to place constant locks on the blocks where the internal (usually directory) information is stored due to concurrent access to them.

Nevertheless, such realization isn't very flexible and is used for tightly coupled networks.

Besides, the system allowed no use of servers with unequal efficiency and connection quality, as data accessibility depended on access to all of the servers, whether they were overloaded or not (with the exception of one with data parity recording).

It simply is not suitable for data storage over the Internet.

Images